What is difference of --prepend-mm-cmds and --append-mm-cmds ?

39 views
Skip to first unread message

人ー

unread,
Dec 9, 2020, 10:16:38 AM12/9/20
to Pantheon
I run the Test.py  with the command shown below.

"python test_flows.py local  -s cubic  --data-dir data_preped --prepend-mm-cmds "mm-delay 50" --extra-mm-link-args "--uplink-queue=droptail --uplink-queue-args=packets=100""

"python test_flows.py local  -s cubic  --data-dir data_append --append-mm-cmds "mm-delay 50" --extra-mm-link-args "--uplink-queue=droptail --uplink-queue-args=packets=100""

When I run it, I got similar throughput and delay time results.

So what's the difference between --prepend-mm-cmds and  --append-mm-cmds ?

人ー

unread,
Dec 9, 2020, 10:51:54 PM12/9/20
to Pantheon
There was a mistake in the command described.

The modified command is shown below.

"python test.py local  -s cubic  --data-dir data_preped --prepend-mm-cmds "mm-delay 50" --extra-mm-link-args "--uplink-queue=droptail --uplink-queue-args=packets=100""

"python test.py local  -s cubic  --data-dir data_append --append-mm-cmds "mm-delay 50" --extra-mm-link-args "--uplink-queue=droptail --uplink-queue-args=packets=100""

Thank you
2020年12月10日木曜日 0:16:38 UTC+9 ー 人:

Keith Winstein

unread,
Dec 9, 2020, 11:17:21 PM12/9/20
to 人ー, Pantheon
Hello -- this is all about constructing a sequence of mahimahi link emulators. Each emulator applies some modeled link effect, e.g. a "delay" or a "loss" or an "onoff" (burst loss) a bottleneck "link" that drains some queue ahead of it.

Sometimes the order matters: for example, it matters a lot whether a "loss" comes before or after a "link". (Because it matters whether a packet is lost before or after it consumes the precious resource of the link capacity.) However, it doesn't matter whether a "delay" comes before or after a link. There's no way to tell whether a packet's propagation delay occurred closer to the sender-side or receiver-side of a queue and link. Therefore, it doesn't really matter whether you "prepend" or "append" the delay emulator to the "link" emulator.

Hope this is helpful!

Best regards,
Keith

--
You received this message because you are subscribed to the Google Groups "Pantheon" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pantheon-stanf...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pantheon-stanford/9f491e76-0b35-4a45-88d6-84f17215ed2en%40googlegroups.com.
Reply all
Reply to author
Forward
0 new messages